On 5/1/23 17:41, Dumitru Ceara wrote:
> On 4/28/23 13:06, Xavier Simonart wrote:
>> Hi Dumitru
>>
> 
> Hi Xavier,
> 
>> When can this "ovn-controller not ready" error happen ? Is this not only at
>> startup?
>> Can this not be fixed using wait_for_ports_up ?
>> This "wait_for_ports_up" is missing for the test failing in ci
>> (../../../tests/ovn.at:26558) - which means that, even with your fix, I am
>> afraid we might still get a flaky test because all flows might not be
>> installed when the packet is injected.
>> It looks to me the missing wait_for_ports_up - which can cause
>> "ovn-controller not ready" or missing flows - happens for these three tests:
>>     - route tables IPv6 -- overlapping subnets
>>     - policy-based routing: 1 HVs, 2 LSs, 1 lport/LS, 1 LR
>>     - policy-based routing IPv6: 1 HVs, 3 LSs, 1 lport/LS, 1 LR
>> if ports are up (after wait_for_ports_up), can the ovn-controller still be
>> not ready?
>>
> 
> I can add these too, for completeness, but AFAICT there are cases when
> "ovn-controller not ready" can happen, e.g., if the address set, port
> group, template incremental processing nodes are in state EN_STALE or
> EN_ABORTED.
> 

We synced up off-list and decided to push this patch as is and follow up
with a patch that adds the missing wait_for_ports_up.  They address
different issues.

I applied the current patch to main and backported it to all stable
branches down to 22.03.

Thanks for the reviews everyone!

Regards,
Dumitru

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to